Re: [Bloat] Achieving Consistent Low Latency for Wireless Real-Time Communications with the Shortest Control Loop

[Zili Meng <zilim@ieee.org>]

[-- Attachment #1: Type: text/plain, Size: 1808 bytes --]

Hi Dave,

Sorry for the late reply. Thank you for your interest in our work! And also
thank you so much for providing the pointers on related literature! They
are really helpful and we will look into these work later.

We did simulate the FqCoDel using the built-in implementation of ns-3. Two
of the issues that we found are that (i) some CCAs are not responsive to
packet losses, especially for latency-sensitive CCAs -- GCC seems to need a
loss rate of >10% to trigger the loss controller; and (ii) if the CCA on
the sender is not ECN-compatible, additional packet retransmissions will
further occupy the buffer space, leading to longer duration of congestion.

We did not compare FqCoDel in the testbed experiments on WiFi AP due to the
reasons above. We do hope to test with more cases and baselines in the real
implementations in the future.

If there are any further questions, please do not hesitate to contact us.
Thank you!

On Sat, Jul 9, 2022 at 6:08 PM Dave Taht <dave.taht@gmail.com> wrote:

> While I deeply appreciate y'all citing our work in this paper,
>
> https://zilimeng.com/papers/zhuge-sigcomm22.pdf
>
> it's not clear if the AP with your solution was actually
> fq_codeled[1]? We've also been battling a ton of regressions of late.
>
> PS It was good to see confirmation of long-tail latency at 99th
> percentile over 400ms in such a large study of gaming users.
>
> PPS Perhaps this paper would be of use?
> https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8469111
>
> [1] https://forum.openwrt.org/t/aql-and-the-ath10k-is-lovely/59002/621
>
> --
> FQ World Domination pending:
> https://blog.cerowrt.org/post/state_of_fq_codel/
> Dave Täht CEO, TekLibre, LLC
>


-- 
Cheers,
----------------
Zili Meng
Web: https://zilimeng.com

[-- Attachment #2: Type: text/html, Size: 3879 bytes --]